Simon McIntosh-Smith Follow @simonmcs
The Micro group runs two Birds of a Feather sessions at IEEE/ACM SuperComputing 2012Micro group staff are running two BOFs at SC'12 in Salt Lake City during November 2012. The first is on Energy Efficient Computing
National HPC resource procurements: ArcherI've been working with EPSRC and NERC as a member of the four person project working group to support the £43m Archer procurement. Archer will be the next national supercomputer resource, replacing HECToR towards the end of 2013. Archer will be several PetaFLOPS in performance and be by far the largest supercomputer in the UK when it is installed.
BBSRC project to fund development of many-core accelerated molecular docking codeAlong with Dr Richard Sessions in Bristol's Biochemistry, we have won a BBSRC grant of ~£150K [BB/K004050/1] to further develop the BUDE molecular docking code. This code already scales across hundreds of GPUs, and is starting early trials with academic and industrial partners. Dr Amaurys Ibarra is the PDRA on this project, starting September 2012.
UoBristol shared infrastructure funding supports new many-core porting roleIn a recent internal competition we won 9 months funding to support the porting of HPC codes developed within the e-Infrastructure South consortium to the Emerald GPU supercomputer. Dan Curran, a recent CS graduate and GPU porting expert, was appointed in July 2012 and is currently working with Prof. Chris Allen to enhance his CFD code so that it can exploit heterogeneous many-core systems, such as GPU accelerated servers.
e-Infrastructure South consortium wins £3.8m to fund HPC Centre of InnovationBristol along with Oxford, Southampton and UCL as formed the e-Infrastructure South consortium. This group won £3.8m in early 2012 to establish a new shared HPC resource. As part of this we have commissioned Emerald, a £1.1m GPU supercomputer, the largest in the UK. For more information on Emerald and other consortium infrastructure, visit the consortium webpages. We plan on making some time on Emerald available to other serious GPU users who'd like to try testing their codes at scale - Emerald has 372 Nvidia M2090 GPUs, along with two high speed interconnects, one using Infiniband, and the other a Gnodal 10GbE network.
Heterogeneous Computing keynote and tutorial at HiPEAC 2012I'm giving the opening keynote as part of the MULTIPROG track at HiPEAC 2012 in Paris (Jan 23-25), as well as co-running a 1-day tutorial on Heterogeneous Computing at the same event (with AMD).
Knowledge Transfer Partnership with Cresset LtdThis project is taking the technologies developed at the University of Bristol for exploiting many-core architectures such as GPUs for molecular (drug) docking and apply them to Cresset's molecular similarity software, FieldScreen. For a brief description of the project, see this blog.
Energy Efficient High Performance Computing website launched with ARM at SC'11 Birds of a Feather sessionThe University of Bristol's Microelectronics Research Group, in partnership with ARM, launched a new community initiative for Energy Efficient HPC with the creation of EEHPC.com. EEHPC.com was announced at an Energy Efficient HPC birds of a feather session at IEEE/ACM SuperComputing 2011 (Nov 15th, Seattle), the first time ARM has officially taken part in the SuperComputing conference. All the talks from the BoF are available from the EEHPC downloads section. Researchers in energy efficient HPC are encouraged to promote their research using this site.
OpenCL Tutorial at SC'11"OpenCL: a hands-on introduction" all-day workshop at IEEE/ACM SuperComputing 2011 (Nov 13th, Seattle) with Tim Mattson (Intel) and Benedict Gaster (AMD). We had roughly 85 particpiants join us for this tutorial.
HPC Software Project: making HPC easy to use for computational chemistryAt the end of 2011 we started a new EPSRC-funded HPC software development project (EP/I030395/1). This project is in collaboration with Adrian Mulholland in the School of Chemistry, and includes two RAs: Dr Christopher Woods and Dr Ben Long. This team is developing a new software library to support massively scalable computational chemistry codes across very large scale systems in a fault-tolerant and energy-aware fashion. The library will be applied to several grand challenge problems in biomolecular simulation. For more information on this project, which is now entering the beta testing stage with early partners, please see the project website.
- Head of Microelectronics Research Group
- Senior Lecturer in High Performance Computing and Architectures
- Member of the EPSRC Archer national supercomputer design team
- Member of the Partnership for Advanced Computing in Europe (PRACE) prototype selection panel for PRACE-1IP (WP9) and PRACE-2IP (WP11), January 2012, allocating a budget of over €2M.
- Invited member of the Scientific Committee for the EPSRC-funded Network on Numerical Algorithms and High Performance Computing, led by Prof Nick Higham at Manchester University.
- Member of the executive board for the EPSRC CCP on Algorithms and Software for Emerging Architectures (ASEArch)
- Chair of the Many-core and Reconfigurable Supercomputing Conference (MRSC)
- Member of EPSRC's new Research Infrastructure Strategic Advisory Team (since 2011)
- Member of EPSRC's resource allocation panel (RAP) for HECToR, the UK's national supercomputer resource
- Member of the programme committee for IEEE/ACM SuperComputing (2012, 2010)
- Part of the International Exascale Software Project since April 2010
- Technical expert in the European Exascale Software Initiative (EESI) since November 2010
- Member of the programme committee for the International Supercomputing Conference
- Member of the programme committee for the Programmability Issues for Heterogeneous Multicores (MULTIPROG) workshop since 2011, part of HiPEAC
- Member of the programme committee for the International Conference on Energy-Aware High Performance Computing since 2010.
- Member of the University of Bristol's HPC executive and technical advisory board
Fifteen years semiconductor industry experience in many-core microprocessor architecture design at Inmos, STMicroelectronics, Pixelfusion and ClearSpeed
- As a founder at ClearSpeed I had various senior technical roles between 2002 and 2008 including VP of Applications and Director of Architecture. I co-developed the architecture of our record-breaking many-core processors, including the 192 core CSX700, 96 core CSX600, and 64 core CS301 prototype
- At Pixelfusion I was a microprocessor architect and low-level GPU driver developer for our innovative 1,536-core graphics processor, the F150. This was the first true GPGPU and also included the world's fastest Rambus implementation at the time (4 channels delivering 6.4 GBytes/s, a lot in 2000!) and the first commercial implementation of an Arc core. It also pioneered using redundant cores to tolerate manufacturing defects in many-core devices.
- Graduated as valedictorian with a first class degree in Computer Science from Cardiff University
- e-Infrastructure South Consortium – Centre for Innovation. [EPSRC EP/K000136/1 & EP/K000144/1]. New regional HPC centre, a consortium of four HPC-intensive universities: Oxford, Bristol, Southampton and UCL. Will create the largest GPU cluster in the UK, and the largest Fermi-based GPU cluster in Europe, to be installed by the end of March 2012, as well as a 12,000 core x86 cluster. ~£3.52M.
- University of Bristol, Faculty of Engineering capital award to purchase a GPU computing teaching cluster, ~£31K.
- Applying many-core Graphics Processors (GPUs) to Molecular Similarity Problems [TSB KTP with Cresset BioMolecular Discovery Ltd]. Beginning Feb 2012, ~£100K.
- Adaptive Multi-Resolution Massively-Multicore Hybrid Dynamics. [EPSRC EP/I030395/1]. Began September 2011, ~£480K.
- Industrial CASE PhD studentship with NAG Ltd. Rob Hunt started Nov 2009. ~£74K.
- New algorithms for novel architectures
- Energy efficient computing
- Heterogeneous, many-core processor architectures, including GPUs, Fusion, MIC, ...
- Scaling applications to run on millions of cores (Exascale computing), including fault-tolerant algorithm design
- Microprocessor 2020
Workshops and Tutorials
- "OpenCL: a hands-on introduction" all-day workshop at IEEE/ACM SuperComputing 2011 (Nov 13th, Seattle) with Tim Mattson (Intel) and Benedict Gaster (AMD). We had roughly 85 participants join us for this tutorial.
- "Introduction to OpenCL" workshop on behalf of EPSRC for users of the UK's national supercomputing resource, HECToR. This was run for the first time in May 2011 and again, in Manchester on November the 24th and 25th, 2011. All the lecture and practical material for this two day workshop is freely available online on my website here. I encourage you to use this for your own purposes, but please retain the appropriate attribution and also drop me a line to let me know!
- I'm also planning an "Advanced OpenCL" workshop for a later date.
- COMSM0109: Advanced Computer Architecture with David May. This course enables you to really understand how processors work and how best to exploit modern processor features from software.
- COMSM2400: High Performance Computing. This course takes you through the theory and hands-on practise of high performance computing, focusing on parallel processing models, and applying these to the latest cutting edge computer architectures.
- COMS12900: HiTec Enterprise (Unit director) - first year introductory course to innovation and entrepreneurship. This course will equip you with highly marketable skills that will prove valuable through your remaining time in Bristol, and beyond.
- COMSM0306: Individual project business plan (Unit director) - fourth year course. Advanced topics on how real companies work, innovation, entrepreneurship and transferrable skills. This is one of the top courses of its kind anywhere in the world (we know, we did the benchmarking), and will make you extremely marketable. The skills you learn here will be incredibly valuable throughout your entire career, no matter what kind of roles you choose.
- Rob Hunt - investigating new mathematical algorithms for massive parallelism. This is a CASE PhD studentship with the Numerical Algorithms Group (NAG)